Deep Learning to Estimate Cardiovascular Risk From Chest Radiographs : A Risk Prediction Study

Abstract

Background: Guidelines for primary prevention of atherosclerotic cardiovascular disease (ASCVD) recommend a risk calculator (ASCVD risk score) to estimate 10-year risk for major adverse cardiovascular events (MACE). Because the necessary inputs are often missing, complementary approaches for opportunistic risk assessment are desirable.

Objective: To develop and test a deep-learning model (CXR CVD-Risk) that estimates 10-year risk for MACE from a routine chest radiograph (CXR) and compare its performance with that of the traditional ASCVD risk score for implications for statin eligibility.

Design: Risk prediction study.

Setting: Outpatients potentially eligible for primary cardiovascular prevention.

Participants: The CXR CVD-Risk model was developed using data from a cancer screening trial. It was externally validated in 8869 outpatients with unknown ASCVD risk because of missing inputs to calculate the ASCVD risk score and in 2132 outpatients with known risk whose ASCVD risk score could be calculated.

Measurements: 10-year MACE predicted by CXR CVD-Risk versus the ASCVD risk score.

Results: Among 8869 outpatients with unknown ASCVD risk, those with a risk of 7.5% or higher as predicted by CXR CVD-Risk had higher 10-year risk for MACE after adjustment for risk factors (adjusted hazard ratio [HR], 1.73 [95% CI, 1.47 to 2.03]). In the additional 2132 outpatients with known ASCVD risk, CXR CVD-Risk predicted MACE beyond the traditional ASCVD risk score (adjusted HR, 1.88 [CI, 1.24 to 2.85]).

Limitation: Retrospective study design using electronic medical records.

Conclusion: On the basis of a single CXR, CXR CVD-Risk predicts 10-year MACE beyond the clinical standard and may help identify individuals at high risk whose ASCVD risk score cannot be calculated because of missing data.

Primary funding source: None.

Figure 1.

Study design. ASCVD=atherosclerotic cardiovascular disease; CVD=cardiovascular disease; CXR=chest radiograph; LDL-C=low-density lipoprotein cholesterol; MACE=major adverse cardiovascular events; MGB=Mass General Brigham; PLCO=Prostate, Lung, Colorectal, and Ovarian. Top. CXR CVD-Risk was developed using data from the PLCO trial. The only input to the model was a CXR image. Independent validation was done in the following 2 groups of participants who were potentially eligible for primary cardiovascular prevention, had no prevalent diabetes or prior MACE, and were not receiving a statin: 8869 outpatients with unknown ASCVD risk due to missing input variables to calculate the traditional ASCVD risk score and 2132 outpatients with known ASCVD risk for whom all input variables to calculate the traditional ASCVD risk score were available. Bottom. The reclassification, calibration, discrimination, and prognostic value of CXR CVD-Risk were evaluated and compared with those of the traditional ASCVD risk score.

Figure 2.

Decision curves for CXR CVD-Risk and the ASCVD risk score. In 8860 outpatients with unknown ASCVD risk (left) and 2132 outpatients with known ASCVD risk (right). ASCVD = atherosclerotic cardiovascular disease; CVD = cardiovascular disease; CXR = chest radiograph.

Figure 3.

Cumulative incidence curves for CXR CVD-Risk and the ASCVD risk score. Cumulative incidence of MACE, by risk category (low, <5%; borderline, ≥5% and <7.5%; intermediate, ≥7.5% and <20%; and high, ≥20%), in the 8869 outpatients with unknown ASCVD risk using CXR CVD-Risk (left) as well as in the 2132 outpatients with known ASCVD risk using CXR CVD-Risk (center) and the traditional ASCVD risk score (right). ASCVD = atherosclerotic cardiovascular disease; CVD = cardiovascular disease; CXR = chest radiograph; MACE = major adverse cardiovascular events.